List your company
ResearchFormerly WizWand

Membership Inference Attack on CIFAR-100 balanced (evaluation set)

99.17AUROCInt. Outs*
53.701265.505677.3189.1144Mar 17, 2022
Updated 8mo ago

Evaluation results

MethodLinks
Int. Outs*Target Model=DenseNet, Requires training=true, Evaluation data used=60%2022.0399.1797.68
Int. Outs*Target Model=ResNet, Requires training=true, Evaluation data used=60%2022.0396.5991.57
Int. Outs*Target Model=ResNext, Requires training=true, Evaluation data used=60%2022.0393.6286.38
Adversarial DistanceTarget Model=ResNext, Requires training=false, Evaluation data used=100%2022.0389.2489.1
WB*Target Model=ResNet, Requires training=true, Evaluation data used=60%2022.0387.5179.73
Adversarial DistanceTarget Model=ResNet, Requires training=false, Evaluation data used=100%2022.0384.5385.45
WB*Target Model=ResNext, Requires training=true, Evaluation data used=60%2022.0384.5276.46
Adversarial DistanceTarget Model=AlexNet, Requires training=false, Evaluation data used=100%2022.0384.3585.12
Adversarial DistanceTarget Model=DenseNet, Requires training=false, Evaluation data used=100%2022.0382.7882.63
WB*Target Model=AlexNet, Requires training=true, Evaluation data used=60%2022.0380.3374.03
WB*Target Model=DenseNet, Requires training=true, Evaluation data used=60%2022.0379.3871.92
Grad w*Target Model=AlexNet, Requires training=true, Evaluation data used=60%2022.0378.7674.32
Grad w*Target Model=ResNext, Requires training=true, Evaluation data used=60%2022.0377.873.47
Grad x*Target Model=ResNext, Requires training=true, Evaluation data used=60%2022.0377.5473.47
Grad x*Target Model=AlexNet, Requires training=true, Evaluation data used=60%2022.0377.273.43
Mentr.Target Model=AlexNet, Requires training=false, Evaluation data used=100%2022.0377.1174.16
Mentr.Target Model=ResNext, Requires training=false, Evaluation data used=100%2022.0376.8775.28
LossTarget Model=AlexNet, Requires training=false, Evaluation data used=100%2022.0376.6974.14
Grad NormTarget Model=AlexNet, Requires training=false, Evaluation data used=100%2022.0376.5874.19
Grad x*Target Model=DenseNet, Requires training=true, Evaluation data used=60%2022.0375.8171.81
Mentr.Target Model=DenseNet, Requires training=false, Evaluation data used=100%2022.0374.2172.69
Grad w*Target Model=DenseNet, Requires training=true, Evaluation data used=60%2022.0373.1272.59
Grad NormTarget Model=ResNext, Requires training=false, Evaluation data used=100%2022.0373.0675.74
LossTarget Model=ResNext, Requires training=false, Evaluation data used=100%2022.0372.5775.17
SoftmaxTarget Model=ResNext, Requires training=false, Evaluation data used=100%2022.0372.3774.84
Grad NormTarget Model=DenseNet, Requires training=false, Evaluation data used=100%2022.0371.373.81
LossTarget Model=DenseNet, Requires training=false, Evaluation data used=100%2022.0370.8572.61
SoftmaxTarget Model=DenseNet, Requires training=false, Evaluation data used=100%2022.0370.5272.11
Grad x*Target Model=ResNet, Requires training=true, Evaluation data used=60%2022.0368.4863.58
SoftmaxTarget Model=AlexNet, Requires training=false, Evaluation data used=100%2022.036865.34
Grad w*Target Model=ResNet, Requires training=true, Evaluation data used=60%2022.0361.9862.72
Grad NormTarget Model=ResNet, Requires training=false, Evaluation data used=100%2022.0359.9362.56
Mentr.Target Model=ResNet, Requires training=false, Evaluation data used=100%2022.0359.161.39
LossTarget Model=ResNet, Requires training=false, Evaluation data used=100%2022.0358.6661.29
Int. Outs*Target Model=AlexNet, Requires training=true, Evaluation data used=60%2022.0357.9256.36
SoftmaxTarget Model=ResNet, Requires training=false, Evaluation data used=100%2022.0355.4557.4